Prognostic Importance of Emotional Support for Elderly Patients Hospitalized With Heart Failure
Background—Several studies have indicated that a variety of social relationships are important predictors of morbidity and mortality in patients with coronary artery disease, but little attention has been focused on the prognostic importance of these factors in the growing population of elderly patients with heart failure. To address this issue, we sought to determine whether emotional support is associated with fatal and nonfatal cardiovascular events in elderly patients hospitalized with heart failure.
Methods and Results—We reviewed the medical records of 292 subjects aged ≥65 years who were hospitalized with clinical heart failure and were part of the New Haven, Conn, cohort of the Established Population for the Epidemiologic Study of the Elderly, a longitudinal, community-based study of aging that included a comprehensive assessment of psychosocial support. In the unadjusted analysis, lack of emotional support was significantly associated with the 1-year risk of fatal and nonfatal cardiovascular outcomes [odds ratio, 2.4; 95% confidence interval, 1.1 to 4.9]. After adjustment for demographic factors, clinical severity, comorbidity and functional status, social ties, and instrumental support, the absence of emotional support remained associated with a significantly higher risk (odds ratio, 3.2; 95% confidence interval, 1.4 to 7.8). The test for interaction between emotional support and sex was significant (P=.01). In the fully adjusted model, the odds ratio for women was 8.2 (95% confidence interval, 2.5 to 27.2) compared with 1.0 (95% confidence interval, 0.3 to 3.3) for men.
Conclusions—Among elderly patients hospitalized with clinical heart failure, the absence of emotional support, measured before admission, is a strong, independent predictor of the occurrence of fatal and nonfatal cardiovascular events in the year after admission. In this cohort, the association is restricted to women.
Heart failure, the most common cause of hospitalization for elderly Medicare patients,1 is associated with high rates of mortality and recurrent cardiovascular events.2 There is great interest in identifying risk factors for adverse outcomes in patients with heart failure to assess prognosis and design interventions to improve outcomes. However, despite advances in risk stratification, there remains much unexplained variation in the outcome of these patients.3
Several studies have indicated that a variety of social relationships are important predictors of morbidity and mortality in patients with coronary artery disease.4 5 However, relatively little attention has been focused on the prognostic importance of these factors in the growing population of elderly patients with heart failure. We were particularly interested in the prognostic importance of emotional support (measure of the presence of intimate contacts) because it had previously been shown to be a prognostic factor in elderly patients with an acute myocardial infarction.6 Accordingly, our specific objective was to determine whether emotional support is associated with fatal and nonfatal cardiovascular events in elderly patients admitted with clinical heart failure. To address this objective, we combined information from a longitudinal, community-based study of aging that included a comprehensive assessment of psychosocial support and a thorough follow-up for adverse events7 with information from hospitalizations for heart failure, based on detailed medical record review.
The study sample population was derived from subjects enrolled in the New Haven, Conn, site of the Established Population for the Epidemiologic Study of the Elderly (EPESE) program.7 This program, which was established in 1982, is a longitudinal, community-based cohort study of 2812 noninstitutionalized men and women aged 65 years and older who are living in New Haven, Conn. The initial cohort was drawn from a probability sample, with stratification by housing type and an oversampling of men.
The study sample for these analyses was restricted to patients in the New Haven EPESE cohort who were hospitalized with heart failure. Throughout the study, the records of the two New Haven hospitals (Yale-New Haven Hospital and the Hospital of St Raphael) were reviewed regularly to identify hospital admissions. We previously determined that >90% of the hospitalizations for this cohort occurred at these two facilities.
The patients enrolled in EPESE who were admitted to the New Haven hospitals with a discharge diagnosis (principal or one of the first three secondary diagnoses) of heart failure (International Classification of Diseases, Clinical Modification, Ninth Revision [ICD-9-CM] code 428) between the beginning of the study in 1982 and December 31, 1992, were identified and their records were reviewed in detail. In addition, Medicare Part A beneficiary data from the Health Care Financing Administration, available since 1985, were matched with EPESE subjects to obtain further information on heart failure admissions. The diagnosis of heart failure in these subjects was confirmed, through medical chart review, by the presence of symptoms of heart failure, and either physical or radiographic findings consistent with heart failure in the first day of admission.
Interviews Before the Hospitalization for Heart Failure
Demographic and Medical Information
Demographic information was collected as part of the initial EPESE interview. For this study, subjects were classified as white or nonwhite. For education, subjects were classified as having less than a high school education, having a high school education, or having more than a high school education. Information about comorbidity (a history of stroke, diabetes, myocardial infarction, hypertension and cancer) was collected based on self-report.
The social and psychological data were collected prospectively in 1982, 1985, and 1988 by trained EPESE interviewers. Annual interviews were conducted in most intervening years but did not include information about psychosocial conditions. The psychosocial data for this study were based on the interview that occurred most directly before the hospitalization for heart failure.
Emotional support was measured by responses to the following question: “Can you count on anyone to provide you with emotional support (talking over problems to help you with a difficult decision)?” The subjects replied “yes,” “no,” or that they had “no need” of emotional support. For the patients who responded “yes,” a follow-up question inquired about the number of sources of support.
Instrumental support was measured using the following question: “When you need some extra help, can you count on anyone to help with daily tasks like grocery shopping, house cleaning, cooking, telephoning, giving you a ride?” The responses were coded as “yes,” “no,” or “no need” of help. For the patients who responded “yes,” a follow-up question inquired about the number of sources of support.
The extent that subjects maintained social connections or ties was assessed at the same time.8 Specifically, subjects were asked about marital status, contact with friends and relatives, membership in religious organizations, and participation in voluntary groups. The measure was used as a count of social ties (coded for this analysis as 0, ≥1).
Depressive symptoms were assessed using the Center for Epidemiologic Study Depressive Symptomatology Questionnaire. This scale has 20 items with a scale ranging from 0 to 60. Based on previous studies, it is a reliable indicator of depressive symptoms in elderly populations.9
Information about functional status was collected from the annual EPESE interview directly preceding the admission for heart failure. Physical functioning was assessed based on self-report items related to basic activities of daily living, physical performance, and gross mobility.10 11 12 These variables were coded as “no impairment” versus “any impairment.”
Medical Record Review
The presence and severity of heart failure were determined from hospital record review. Clinical characteristics included heart rate, mean blood pressure, respiratory rate, temperature, hematocrit, white blood cell count, creatinine, sodium, and potassium. These variables were combined in an acute physiology score, based on the work of Knaus et al.13 A variable was created to indicate whether the index admission was associated with an acute myocardial infarction. Comorbid conditions also were based on the medical chart review and the EPESE interview preceding the hospitalization.
The principal outcome of this study was the occurrence of any fatal or nonfatal cardiovascular event in the year after hospital admission with heart failure. Since the inception of the cohort, <1% of the subjects have been lost to follow-up. Deaths or hospitalizations due to cardiovascular disease were defined by ICD-9-CM codes (391–398, 402, 404, 410–416, and 420–429) on the death certificate or as the principal discharge diagnosis. The death certificates were coded by a certified nosologist unfamiliar with the study hypothesis.
We sought to determine the bivariate association of emotional support and other potential covariates with the outcome variable of cardiovascular event within 1 year of admission. We also evaluated the association between patient characteristics and emotional support. Significant associations were identified using the χ2 statistic. We determined cardiovascular events by level of emotional support within 30, 60, 90, 180, and 365 days after admission.
We conducted the multivariable analysis using two approaches to evaluate the independent association of emotional support with the occurrence of cardiovascular events. First, using the variables from the bivariate analysis, we developed a multivariable logistic regression model with stepwise selection to predict the occurrence of a cardiovascular event within 1 year after admission, using an exit level of P<.10. Age was forced into the model at each step. For the multivariable analysis, we selected logistic regression because visual inspection of event curves suggested that they did not satisfy the assumption of proportionality that is required for the use of proportional hazards models.
Next, we more formally tested the association of emotional support with the outcome of cardiovascular events within 1 year of admission with a series of hierarchical logistic regression models. In the first model, we included emotional support (no support compared with any support; an indicator variable was used for patients who responded that they had no need for emotional support) as the sole explanatory variable. In the second model, we included emotional support and demographic variables (age, sex, and race) as explanatory variables. In the third model, we added clinical severity variables (ejection fraction, acute physiology score [quartiles], history of myocardial infarction, and current myocardial infarction) that were associated with the outcome. In the fourth model, we added comorbidity (presence of stroke, diabetes, chronic obstructive pulmonary disease, dementia, chronic renal failure, or tumor) and functional status variables (limitations in activities of daily living or gross mobility). Finally, we added two other psychosocial factors (instrumental support and social ties).
Using the final model, we checked for interactions between emotional support and sex, age, ejection fraction, and presence of an acute myocardial infarction. We also used the same hierarchical models to evaluate the association between emotional support and the secondary outcome variables.
We repeated the analysis in several ways to be certain that the results were not dependent on any single factor concerning the assembly of the study sample or the definition of the variable. First, we repeated the final model after excluding patients who had been admitted with an acute myocardial infarction to ensure that the result was not dependent on this group. We also repeated the model after combining patients who stated that they did not need emotional support with those who explicitly stated that they did not have any emotional support.
For all models, we constructed and examined partial residual plots to evaluate potential problematic areas of model fit.14 Goodness-of-fit was evaluated by comparing fitted probabilities of 365-day cardiovascular mortality or readmission with observed 365-day mortality within deciles of risk and calculating the corresponding observed χ2 statistics.15
A total of 473 hospital admissions were screened, with 295 confirmed with heart failure. Of these patients, 3 were excluded because of missing mortality data (2 with missing cause of death and 1 with missing death date). The psychosocial interview was conducted 679±517 days before the index hospitalization.
At the time of admission, most of the 292 patients were between 75 and 84 years old (mean age, 80.0±7.2 years), and 57% were female. About half of the group had a history of heart failure, and 40% had a history of a myocardial infarction. Before the index hospitalization, functional impairment was common, occurring in 80% of the subjects.
A total of 38 subjects (13%) reported no sources of emotional support, 48 (16%) reported no need of emotional support, and 206 (71%) had one or more sources of support. We evaluated the stability of this measurement. Among the 149 patients in our study sample who had at least two psychosocial interviews before the index hospitalization, about one fifth changed categories (19.5% for those who subsequently had a cardiovascular event and 19.4% for those who did not). Emotional support was not significantly associated with clinical characteristics, including the acute physiology score, history of heart failure, myocardial infarction during the index hospitalization, ejection fraction, diabetes, hypertension, or the number of comorbid conditions.
In the 1 year after hospital admission, 142 patients (49%) experienced a cardiovascular death or readmission. A total of 75 patients were readmitted at least once in the year after discharge. A hospitalization for heart failure was the most common reason for readmission (48% of the cases), followed by coronary artery disease (8%) and unstable angina (7%). Overall, 110 of the patients died during the year after admission, including 83 from cardiovascular causes.
In the bivariate analysis, there were several factors that were associated with an increased risk of cardiovascular events in the year after admission (Tables 1⇓ and 2⇓). Men were more likely to have an event than women (P=.04). Patients with systolic dysfunction had a higher risk than patients with normal systolic function (P=.05). Patients who had a myocardial infarction on the index admission had a much higher risk of subsequent events (P=.004). A worse acute physiology score was also associated with an increased risk of cardiovascular events, although the significance was borderline (P=.06 for trend).
Patients who reported no source of support had a higher likelihood of a cardiovascular event within 1 year of admission (P=.02) compared with those who had sources of support. The association between the absence of emotional support and the risk of cardiovascular events, not present within the first 30 days after admission, became prominent over time (Table 3⇓).
Patients with no social ties had a higher rate of events than those with ties, but the result was borderline significant (P=.1). There was no strong association between depressive symptoms or instrumental support and the occurrence of cardiovascular events.
Figs 1 to 4⇓⇓⇓⇓ show the risk of cardiovascular events after stratification by age, sex, ejection fraction, and acute myocardial infarction. Except for the stratification by sex, the association between the lack of emotional support and the risk of cardiovascular events was evident in each of the subsets. However, in the sex analysis, the association of emotional support and cardiovascular events at 1 year appeared to be restricted to women.
The stepwise multiple logistic regression model predicting the occurrence of cardiovascular events in the year after admission is shown in Table 4⇓. In this model, lack of emotional support was associated with a significant odds increase in risk (odds ratio [OR], 2.7; 95% confidence interval [CI], 1.2 to 6.0). The absence of social ties was also borderline significant in predicting events (OR, 2.1; 95% CI, 0.95 to 4.5). No other psychosocial factor entered the model.
The most important clinical variables that predicted cardiovascular events were the presence of an acute myocardial infarction on the index admission (OR, 2.2; 95% CI, 1.2 to 4.0), an abnormal (OR, 1.6; 95% CI, 0.7 to 3.6) or missing (OR, 2.5; 95% CI, 1.2 to 5.1) left ventricular ejection fraction, and a worse acute physiology score (OR, 1.7; 95% CI, 0.94 to 3.2). The only comorbidity that was independently predictive of an event was a history of hypertension (OR, 2.0; 95% CI, 1.1 to 3.6). Functional impairment before admission (OR, 1.7; 95% CI, 0.9 to 3.3) and male sex (OR, 2.0; 95% CI, 1.1 to 3.4) also entered the model. Age was forced in the model but was not significantly associated with the outcome (OR, per year 1.00; 95% CI, 0.96 to 1.04).
The test for interaction between emotional support and sex was significant (P=.01). In the fully adjusted model, the OR for women was 8.2 (95% CI, 2.5 to 27.2) compared with 1.0 (95% CI, 0.3 to 3.3) for men. Neither age, myocardial infarction during the index hospitalization, nor ejection fraction had a significant interaction with emotional support.
We repeated the analysis with several modifications to examine the stability of our result. With the addition of covariates that could have been potential confounders, the association became stronger. In the final model, after adjustment for demographic factors, clinical severity, comorbidity and functional status, social ties, and instrumental support, the absence of emotional support had a significant association with cardiovascular events (OR, 3.2; 95% CI, 1.4 to 7.8). The absence of emotional support remained an important predictor of cardiovascular events if the patients who expressed no need of emotional support were combined with those who reported no source of support (OR, 2.6; 95% CI, 1.2 to 5.8); the sample was restricted to the 214 patients who did not experience a myocardial infarction (OR, 2.8; 95% CI, 1.1 to 7.5); the model was adjusted for the number of days between the assessment of emotional support and the index admission (OR, 3.3; 95% CI, 1.4 to 7.9); or the outcome was cardiovascular deaths (OR, 2.6; 95% CI, 1.0 to 6.6), cardiovascular readmissions (OR, 2.5; 95% CI, 0.9 to 7.1), or all-cause deaths (OR, 2.0; 95% CI, 0.8 to 5.0). Emotional support was not a predictor of the 27 noncardiovascular deaths.
Various aspects of social relationships are being examined with increasing frequency as potential risk factors for adverse outcomes among patients with coronary heart disease. We make several observations that advance this field of inquiry. First, our principal finding is that the absence of emotional support is a strong predictor of cardiovascular events in elderly patients admitted to the hospital with heart failure. Patients without emotional support had a more-than-threefold odds increase in the risk of cardiovascular events in the year after admission compared with patients with emotional support.
Second, this association was not explained by differences in clinical severity on admission or a susceptibility to early adverse outcomes. In fact, the converse was observed. Patients who reported no source of emotional support tended to be admitted with less clinically severe disease and the association with adverse outcomes was not apparent until several months after discharge. Thus, it is unlikely that disease severity influenced the availability of emotional support.
Third, a commonly proposed mechanism that links lack of emotional support with poor prognosis is related to the ability of patients to obtain tangible assistance in getting to physician appointments and maintaining independent living (eg, obtaining groceries, cooking). Although this is a difficult hypothesis to test, our data indicate that neither instrumental support nor functional status mediates the relationship between emotional support and adverse outcomes. In fact, instrumental support (eg, getting help or assistance with activities like grocery shopping and transportation to medical appointments) is not a significant prognostic factor.
This study complements the previous investigation from this cohort that established the prognostic importance of emotional support for patients with an acute myocardial infarction.6 Not only do we extend the observation to patients with heart failure but we also specifically evaluate the association with cardiovascular events and evaluate longer-term follow-up. However, because we included all patients with clinical heart failure, almost a fourth of our study sample had an acute myocardial infarction during the index hospitalization. The redundancy in study samples was minimized because of the different time frame of the two studies, with 25% of the patients in both study samples. Nevertheless, the results do not change when patients with an acute myocardial infarction are excluded from our study sample.
A provocative finding in this study is that the association between emotional support and cardiovascular events was very strong in women but absent in men. The test for an interaction in the fully adjusted model was significant with a value of P=.01, a striking difference worthy of further investigation. Interestingly, in the study of the prognostic importance of emotional support after acute myocardial infarction, there was no interaction with sex.6
Other Psychosocial Factors
Previous studies have evaluated a variety of psychosocial factors. Although others, focusing on patients with coronary heart disease, have examined the effect of living alone16 or not being married as risk factors,17 our study also highlights the importance of social ties as well as emotional support from intimate contacts. Although not the primary focus of this study, the absence of social ties did demonstrate a more-than-twofold odds increase in risk of cardiovascular events that was independent of emotional support. This factor has been associated with all-cause mortality,18 but its relationship with cardiovascular events in patients hospitalized with heart failure has not been previously described.
In contrast, although other studies have focused on the prognostic importance of depression in other patient populations,19 20 21 we did not find depressive symptoms to be independently associated with cardiovascular events. Of note, our measure of depression was based on symptoms, rather than a clinical diagnosis, and the assessment preceded the hospitalization.
Why do subjects who lack emotional support have higher rates of cardiovascular events? The mechanism is not known. It is possible that emotional support is associated with greater patient adherence to medical therapy and lifestyle recommendations. Although an evaluation of such a mechanism is beyond the scope of this study, it is relevant to note that the association between emotional support and cardiovascular outcomes did not appear to be mediated by smoking, obesity, or physical activity. Another possibility is that the effect of emotional support is mediated through a direct physiological pathway. For example, the emotional support may mitigate potentially damaging effects of negative emotional interactions on neuroendocrine and physiological regulatory systems.18 22 23
Whatever the mechanism, it is interesting that it becomes manifest only months after hospital admission. Consequently, the absence of psychosocial support may be a good target for intervention in these patients. This approach is gaining momentum in the treatment of patients with coronary artery disease. In a meta-analysis of 23 randomized controlled trials of psychosocial intervention in the setting of cardiac rehabilitation in patients with coronary artery disease, the addition of psychosocial treatments was found to decrease mortality, morbidity, and psychological distress and improve risk factors.24 The National Institutes of Health is sponsoring the Enhancing Recovery in Coronary Heart Disease (ENRICHD) Patients Study, a clinical trial to evaluate the effect of a psychosocial intervention for depressed and/or socially isolated patients hospitalized with an acute myocardial infarction.
Elderly patients with heart failure may gain the most benefit from these interventions because their event rates are so high and there are relatively few therapeutic options available to them. Recently, a trial of a multidisciplinary intervention with psychosocial components for patients with heart failure demonstrated a reduction in readmission within 90 days of discharge from 42.1% to 28.9%.25 The cost of the intervention was more than offset by the saving from the readmissions that were avoided.
This study has several limitations. The timing of the assessment of support occurred a mean of almost 2 years before the admission with heart failure. As a result, we may have misclassified some patients based on their status at the time of admission. In addition, the assessment was made with a single-item instrument, and there was no formal assessment of its reliability. These limitations, however, would tend to bias the study toward the null hypothesis. Despite such limitations in our measurement, we observed a strong association of emotional support with cardiovascular outcomes.
Our study identifies emotional support as an important prognostic factor for elderly patients hospitalized with heart failure, especially in women. The association, not present in the first 30 days after admission, grows in strength over time and is independent of demographic, clinical, and other psychosocial factors. Future studies are necessary to illuminate the mechanism of this association and explore the value of interventions designed to augment sources of emotional support.
This work received the following support from the National Institute on Aging: Contract #N01-AG-0-2105 (EPESE), Grant #P60-AG-10469 (PEPPER). Dr Krumholz is a Paul Beeson Faculty Scholar.
Reprint requests to Harlan M. Krumholz, MD, Yale University School of Medicine, 333 Cedar St, PO Box 208025, New Haven, CT 06520-8025.
- Received July 21, 1997.
- Revision received October 14, 1997.
- Accepted November 19, 1997.
- Copyright © 1998 by American Heart Association
Hennen J, Krumholz HM, Radford MJ. Twenty most frequent DRG groups among Medicare inpatients age 65 or older in Connecticut hospitals fiscal years 1991, 1992, and 1993. CT Med. 1995;59:11–15.
Berkman LF. The role of social relations in health promotion. Psychosom Med. 1995;57:245–254.
Berkman LF, Leo-Summers L, Horwitz RI. Emotional support and survival after myocardial infarction. Ann Intern Med. 1992;117:1003–1009.
Berkman LF, Berkman CS, Kasl SV, Freeman DJ Jr, Leo L, Ostfeld AM, Cornoni-Huntley J, Brody JA. Depressive symptoms in relation to physical health and functioning in the elderly. Am J Epidemiol. 1986;124:372–388.
Katz S, Downs TD, Cash HR, Grotz RC. Progress in the development of the index of ADL. Gerontologist. 1970;10:20–30.
Rosow I, Breslau N. A Guttman health scale for the aged. J Geront. 1966;21:556–559.
Landwehr JM, Pregibon D, Shoemaker AC. Graphical methods for assessing logistic regression models. JASA. 1984;79:61–63.
Hosmer DW, Lemeshow S. Applied Logistic Regression. New York, NY: Wiley; 1989.
Chandra V, Szklo M, Goldberg R, Tonascia J. The impact of marital status on survival after an acute myocardial infarction: a population based study. Am J Epidemiol. 1983;117:320–325.
Seeman TE, Berkman LF, Blazer DG, Rowe JW. Social ties and support and neuroendocrine function. Ann Behav Med. 1994;16:95–106.
Frasure-Smith N, Lespérance F, Talajic M. Depression and 18-month prognosis following myocardial infarction. Circulation. 1995;91:999–1005.
Cohen S. Psychosocial models of the role of social support in the etiology of physical disease. Health Psych. 1988;7:269–297.
Gerin W, Milner D, Chawla S, Pickering TG. Social support as a moderator of cardiovascular reactivity in women: a test of the direct effects and buffering hypotheses. Psychosom Med. 1995;57:16–22.